Blood or buzz? Decoding the quality of CBC-related short videos on Chinese platforms

Abstract

Background

The complete cell count (CBC) is a fundamental diagnostic tool in clinical practice and is essential for screening and managing diseases such as anemia, infections, and malignant blood disorders. In China, a rapidly aging population and a growing burden of chronic diseases have increased the demand for accessible health knowledge resources. Short videos have now become a popular channel for medical information dissemination. Therefore, this study aimed to assess the overall quality and credibility of videos about CBC in China.

Objective

The aim of this study was to assess the information quality of CBC-related videos on short video sharing platforms in China.

Methods

We searched for short videos that popularize the main knowledge of CBC posted on three short video platforms in China that are currently accessed with a large amount of information: Douyin, Bilibili, and Rednote. A total of 242 relevant videos were retrieved, and we collected, processed, and analyzed the authors and basic information of all videos. The quality and reliability of their contents were assessed by using the Global Quality Score scale, the Modified DISCERN Medical Video Quality Evaluation Tool. Subsequently, short video platforms as well as video publishers were analyzed and compared descriptively as a whole. Potential correlations between general video information and video quality and reliability were analyzed by Spearman's correlation coefficient.

Results

The quality of online videos provided by short video platforms showed a moderate level of quality (just 49.5% met the high-quality level criterion), and the completeness of their content, as well as their reliability was average (only 28.9% of videos met the reliability criterion). Further results on group comparisons showed that videos from healthcare professionals were better than those from non-healthcare professionals in terms of content comprehensiveness, reliability, and quality. Additionally, we observed a positive correlation between the production length of the video and video quality; however, there was no significant correlation between video likes and comments, etc., and video quality.

Conclusion

The results of this study suggest that the overall quality and reliability of short videos about CBC-related content in current short video platforms are still significantly deficient. It is recommended that viewers should treat this content with caution. Among them, videos posted from medical personnel are more instructive. Nevertheless, video-based popularization of medical knowledge still holds promise. The overall quality and reliability of medical information shared on short video platforms can be improved by implementing appropriate strategies.

Introduction

The complete blood count (CBC) is a foundational diagnostic tool in clinical medicine, providing a comprehensive snapshot of hematologic health through quantitative analysis of erythrocytes, leukocytes, and platelets. ¹ Globally, CBC testing is integral to screening, diagnosis, and monitoring of conditions ranging from infections ² and anemia ³ to hematologic malignancies. ⁴ In China, the clinical significance of CBC is amplified by rapid demographic shifts: the proportion of individuals aged ≥60 years is projected to rise from 12.4% (168 million) in 2010 to 28% (402 million) by 2040, ⁵ accompanied by a growing burden of chronic noncommunicable diseases (NCDs), which account for nearly 80% of deaths in this population. ⁶ Complete cell count serves as a frontline tool for managing NCDs such as diabetes and hypertension,^7,8 enabling early detection of complications such as anemia of chronic disease and guiding therapeutic adjustments in chemotherapy^9,10 or postoperative care.^11,12 Despite the extensive clinical utilization of CBC testing, studies indicate that Chinese patients struggle to interpret common lab indicators and may misjudge disease risks. ¹³ This difficulty is worsened by limited patient–provider communication in China's high-volume healthcare system. ¹⁴ This underscores the urgent need for accessible, evidence-based educational resources to bridge health literacy disparities.

The rapid expansion of internet-based health resources has transformed how individuals seek medical information. ¹⁵ In China, platforms such as Douyin, Bilibili, and Rednote have emerged as dominant channels for health communication, combining engaging video formats with algorithms that prioritize user-friendly content. As early as 2006, China surpassed the United States to become the world's largest number of Internet users. ¹⁶ According to the 53rd Statistical Report on Internet Development in China released by the China Internet Network Information Center, the country's internet user base reached 1.092 billion by December 2023, with a penetration rate of 77.5%. ¹⁷ Increasingly, these users rely on social media platforms for health-related guidance. Compared to traditional text-based materials, video platforms hold distinct advantages in conveying complex medical concepts through visual storytelling, which enhances comprehension and retention. Bilibili, favored by younger demographics, emphasizes long-form, community-driven medical explanations, often featuring annotated laboratory workflows, while Douyin's algorithm prioritizes brevity, amplifying concise “health hacks.” ¹⁸ Rednote, popular among young consumers for its integration of lifestyle content with patient narratives, ¹⁹ has seen its content ecosystem further enriched by recent shifts in the global digital landscape, driving its growing prominence. However, the democratization of content creation carries significant risks: unverified or misleading information, often produced by nonexperts, can propagate rapidly, ²⁰ exacerbating health anxieties or encouraging inappropriate self-management.

Although extensive research has been conducted on Chinese online health content for cancers such as laryngeal carcinoma, ²¹ liver cancer, ²² gastric cancer, ²³ breast cancer, ²⁴ and other conditions such as migraine ²⁵ and nonalcoholic fatty liver disease, ²⁶ no studies have yet evaluated or analyzed the quality of short videos related to laboratory testing. Complete cell count–related materials remain understudied—a significant oversight given their critical role in preventive healthcare. In China, the unique mechanisms of short video platforms present distinct challenges. For instance, Bilibili's “coin” reward system incentivizes attention-grabbing and exaggerated titles, while Douyin prioritizes immediate audiovisual impact, resulting in “fast-consumption” content that fails to ensure accurate or reasonable explanations of medical topics. Douyin's lifestyle-focused communities often conflate anecdotal experiences with medical advice. These dynamics highlight the urgent need for platform-specific quality assessments, particularly under China's Healthy China 2030 framework, which prioritizes expanding health education through new media. ²⁷

In light of this, our research team conducted searches and collected data from three of the most popular local video platforms in this study. Through systematic analysis of metadata, engagement metrics, and content validity, we aimed to investigate the overall quality and reliability of CBC-related online video resources in mainland China. These findings provide actionable strategies for policymakers, healthcare providers, and platform developers to improve the quality of CBC-related health communication in the digital age.

Search strategy and data collection

In this cross-sectional study examining Chinese videos, we systematically retrieved publicly accessible content related to CBC testing from three major platforms: Douyin (China's Douyin, http://www.douyin.com), BiliBili (http://www.bilibili.com), and Rednote (http://www.Douyin.com). Detailed descriptions of these platforms, including their operational dynamics and user engagement features, are provided in Supplementary Table 1.

Data collection was conducted within a single day (March 13, 2025) in Tianjin, China, using a Mac computer (macOS Sequoia 15.1.1) equipped with a newly installed Safari browser (version 18.1.1). To mitigate algorithmic bias, all cookies, cached data, browsing history, and autofill information were systematically purged prior to initiating searches. For Rednote, which mandates user login for search functionality, a new account was registered using an unused mobile number to eliminate residual user data influences and ensure standardized access.

The search term “血常规” (“Complete Blood Count” in Chinese) was employed to retrieve the top 100 default-ranked videos per platform. Prior studies have validated that content beyond the top 100 videos on such platforms exerts negligible influence on analytical outcomes, as these rankings adequately capture the breadth of relevant content within the domain.^23,26 Initial exploratory analyses identified a subset of CBC-related videos addressing veterinary applications (e.g., blood tests for companion animals), likely reflecting rising public interest in animal welfare in China.^28,29 These were excluded to maintain the study's focus on human healthcare. Consequently, exclusion criteria were applied to refine the dataset: (1) non-Chinese language content (including dialects); (2) irrelevance to human CBC testing (e.g., veterinary use, unrelated medical topics); (3) payment is required; (4) duplicate uploads within the same platform.

A total of 300 videos (100 per platform) were initially identified. After excluding 1 non-Chinese videos, 31 duplicates, 23 irrelevant entries, and 3 videos that require payment, 242 videos were retained for data extraction and content evaluation. For each video, metadata such as URLs, publication date, video length, and engagement metrics (likes, shares, saves, comments) were recorded. All data were stored in WPS Office spreadsheets for consistency and accessibility. A detailed flowchart outlining the screening process is presented in Figure 1.

Figure 1.

The flowchart of search strategy, video screening procedure, and further analysis of the study.

Ethical considerations

This study utilized publicly available video data without accessing private user information or clinical records. Ethical approval was not required, as the research involved no human participants or sensitive data.

Classification of videos

Based on the identity of the poster, short videos are categorized into two major groups: (1) medical practitioners and (2) nonmedical practitioners. Medical practitioners are further subdivided into: (1) hematologists, (2) nonhematology clinical physicians, (3) practicing laboratory physicians, and (4) medical students; Nonmedical practitioners are divided into: (1) science communicators and (2) patients, which are provided in Supplementary Table 2.

Based on the video presentation design, we categorize it into three aspects: background environment, visual communication design, and visual text. The background environment is further divided into: (1) clinical laboratory, (2) other environments within the hospital, (3) outside the hospital, and (4) virtual backgrounds. The visual communication design is classified into: (1) monologue (without diagram demonstration), (2) audio 2D/3D animation, and (3) solo narration (with diagram demonstration). Additionally, the visual text is categorized into: (1) subtitled and (2) nonsubtitled, which are provided in Supplementary Table 3.

The video content is classified across two dimensions. One dimension involves the specific indicators of a CBC, which encompasses 30 common parameters such as white cell count (WBC), red cell count, hemoglobin (HGB), and others listed in detail in Supplementary Table 4. The other dimension pertains to the diseases covered in the videos, which include Infectious Diseases, Anemia, Coagulation Disorders, Allergic Diseases, Tumors, conditions Related to Radiotherapy and Chemotherapy, Immune System Disorders, and Hematologic Diseases. If a video provides an interpretation of the aforementioned laboratory indicators in the context of these specific diseases, it will be documented.

Video assessment

Two qualified researchers with adequate medical education and training backgrounds (Gao Ying and Xu Xinyi) independently scored the short videos using three evaluation tools: Modified DISCERN (mDISCERN), Global Quality Score (GQS), and Medical Quality Video Evaluation Tool (MQ-VET) scoring standard, with detailed evaluation methods provided in Supplementary Tables 5–8. In case of disagreement, it should be resolved through discussion or negotiation with the third researcher (Jianbo Xu). Before the official scoring, the three raters had fully reviewed and understood the questionnaires in detail to minimize potential biases arising from misunderstandings of the scoring tools.

Statistical analyses

For continuous variables, the Shapiro–Wilk test was used to assess their distribution characteristics. The results indicated that all continuous variables did not follow a normal distribution. Therefore, descriptive statistics were presented as medians (interquartile ranges). The Kruskal–Wallis nonparametric test was employed for between-group comparisons. Categorical variables were analyzed using the chi-square test or Fisher's exact test to evaluate their significance. Spearman's correlation analysis was conducted to examine the relationship between frequency quality assessment indicators, reliability parameters, and basic characteristics. All analyses were performed in R (version 4.2.3), primarily utilizing the gtsummary package (version 2.0.4) for data summarization and statistical inference. Hypothesis testing was conducted using two-sided tests, with a significance level set at α = 0.05. P < 0.05 was considered statistically significant.

Results

General information of eligible videos

This study conducted data analysis on 242 videos that met the inclusion criteria (Table 1). Among the total sample, the median video duration was 72 s (IQR: 32.0–150.75), the median number of likes was 137 (IQR: 25.25–1580.75), the median number of comments was 7 (IQR: 1.0–133.75), and the median number of shares was 194 (IQR: 26.0–1258.0). Significant differences were observed in the indicators among different platforms (all p < 0.001). Douyin had the shortest video duration (median of 57 s), but its interaction indicators (758 likes, 159 comments, 435 shares) were significantly higher than those of other platforms.

Table 1.

Baseline characteristics of the videos.

	Total	BiliBili	Douyin	Rednote	p value
General information
Number (n)	242	64	89	89
Length (s, median, IQR)	72.000 (32.000;150.750)	704.000 (139.250;1571.000)	57.000 (31.000;94.000)	40.000 (22.000;75.000)	<0.001
Days since published (n, median, IQR)	148.500 (30.500;429.000)	440.000 (95.000;845.250)	63.000 (23.000;227.000)	165.000 (41.000;389.000)	<0.001
Likes (n, median, IQR)	137.000 (25.250;1580.750)	85.000 (23.750;1412.000)	758.000 (138.000;5831.000)	40.000 (10.000;183.000)	<0.001
Saves (n, median, IQR)	40.000 (6.250;348.750)	146.500 (33.750;2021.000)	29.000 (3.000;156.000)	32.000 (5.000;183.000)	<0.001
Comments (n, median, IQR)	7.000 (1.000;133.750)	1.000 (0.000;13.250)	159.000 (14.000;2662.000)	2.000 (0.000;20.000)	<0.001
Shares (n, median, IQR)	194.000 (26.000;1258.000)	42.000 (7.000;329.000)	435.000 (84.000;5639.000)	N/Aa	<0.001
Publisher (n, %)					.
Medical practitioners
Nonhematology clinical physicians	113 (46.694%)	18 (28.125%)	56 (62.921%)	39 (43.820%)
Hematologists	51 (21.074%)	4 (6.250%)	29 (32.584%)	18 (20.225%)
Practicing laboratory doctors	9 (3.719%)	4 (6.250%)	1 (1.124%)	4 (4.494%)
Medical students	5 (2.066%)	5 (7.812%)	0 (0.000%)	0 (0.000%)
Nonmedical practitioners
Science communicator	53 (21.901%)	33 (51.562%)	1 (1.124%)	19 (21.348%)
Patients	11 (4.545%)	0 (0.000%)	2 (2.247%)	9 (10.112%)
Background environment (n, %)					<0.001
Clinical laboratory	10 (4.132%)	2 (3.125%)	0 (0.000%)	8 (8.989%)
Virtual backgrounds	137 (56.612%)	48 (75.000%)	30 (33.708%)	59 (66.292%)
Other environments within the hospital	63 (26.033%)	3 (4.688%)	45 (50.562%)	15 (16.854%)
Outside the hospital	32 (13.223%)	11 (17.188%)	14 (15.730%)	7 (7.865%)
Visual communication design (n, %)					<0.001
Monologue	10 (4.132%)	4 (6.250%)	6 (6.742%)	0 (0.000%)
Audio 2d/3d animation	180 (74.380%)	46 (71.875%)	75 (84.270%)	59 (66.292%)
Solo narration	52 (21.488%)	14 (21.875%)	8 (8.989%)	30 (33.708%)
Visual text (n, %)					<0.001
Nonsubtitled	56 (23.140%)	33 (51.562%)	3 (3.371%)	20 (22.472%)
Subtitled	186 (76.860%)	31 (48.438%)	86 (96.629%)	69 (77.528%)
Index Integrity, n (%):					0.316
Fully reported	113 (46.694%)	30 (46.875%)	47 (52.809%)	36 (40.449%)
Not mentioned	21 (8.678%)	7 (10.938%)	4 (4.494%)	10 (11.236%)
Partially Reported	108 (44.628%)	27 (42.188%)	38 (42.697%)	43 (48.315%)
Quality and reliability
GQS score, Median (Q1–Q3)	3.000 (3.000;4.000)	4.000 (3.000;4.000)	4.000 (4.000;4.000)	3.000 (2.000;3.000)	<0.001
mDISCERN, Median (Q1–Q3)	3.000 (2.000;4.000)	3.000 (3.000;4.000)	3.000 (3.000;4.000)	2.000 (2.000;3.000)	<0.001
MQ-VET score, Median (Q1–Q3)	58.000 (52.000;62.000)	60.500 (56.000;65.000)	61.000 (57.000;63.000)	53.000 (47.000;56.000)	<0.001

Regarding the identity of the video creators, 45.9% of the videos were created by Non-Hematology Clinical Physicians, with Douyin having the highest proportion (62.9%). Videos made by science communicators accounted for 21.9%, mainly concentrated on Bilibili (51.6%). Patient-posted videos accounted for 4.5%, primarily from Douyin (10.1%).

In terms of video production, 56.6% of the videos used virtual backgrounds (75.0% on Bilibili), while 50.6% of Douyin videos featured hospital interiors as backgrounds; 74.4% of the videos adopted 2D/3D animation, with Douyin having the highest proportion (84.3%); 96.6% of Douyin videos included subtitles, significantly higher than other platforms. Regarding indicator completeness, 46.7% of the videos contained all indicators, with no significant difference among platforms (p = 0.316).

Information quality and reliability

This study systematically evaluated the quality and reliability of blood routine science videos on various platforms using the GQS, mDISCERN, and MQ-VET (Table 1, Figure 2). The overall sample analysis indicated that the video quality and reliability were at a moderate level, with median GQS scores of 3.00 (IQR: 3.00–4.00), median mDISCERN scores of 3.00 (IQR: 2.00–4.00), and median MQ-VET scores of 58.00 (IQR: 52.00–62.00). The comparison among platforms showed that Rednote had lower median scores in all evaluation dimensions compared to other video platforms, with all p < 0.05 (Figure 2(a)–(f)). Notably, videos produced by nonmedical practitioners and published on Bilibili had significantly higher median scores than those on other platforms (p < 0.05), while there was no statistically significant difference in video quality between Rednote and Douyin platforms.

Figure 2.

The differences in video quality and reliability among specific video-sharing platforms. (A–C) Overall. (D–F) Medical practitioners. (G–I) Nonmedical practitioners. *p < 0.05; **p < 0.01; ***p < 0.001.

A comparative analysis based on the professional backgrounds of the publishers revealed that videos created by medical practitioners significantly excelled in GQS, mDISCERN, and MQ-VET scores compared to those made by nonmedical practitioners (with all p < 0.05; Figure 3(a)–(c)). The results of multiple comparisons between groups further indicated that: in the GQS scoring dimension, videos published by patients received significantly lower scores than those by Non-Hematology Clinical Physicians, hematologists, Practicing laboratory doctors, and Science communicators (p < 0.05). In the mDISCERN scoring dimension, the patient group's scores were notably lower than those of nonhematology clinical physicians, hematologists, Medical Students, and Science communicators (p < 0.05). Regarding the MQ-VET scoring dimension, videos produced by patients demonstrated the lowest scoring trend among all professional groups, with statistically significant differences compared to the other five groups (with all p < 0.05).

Figure 3.

The differences in video quality and reliability among video producer identity. (A–C) Medical practitioners and nonmedical practitioners. (D–F) Various publishers. *p < 0.05; **p < 0.01; ***p < 0.001.

Subgroup analysis of platforms (Supplementary Tables 9–11) found no statistically significant difference in video quality between clinicians and nonclinical professionals on the Bilibili platform (p > 0.05). However, on Rednote and Douyin platforms, videos made by clinicians were significantly better than those made by nonclinical professionals in all evaluation dimensions (p < 0.05).

Correlation analysis

This study employed heatmap visualization to analyze the correlations among three quality evaluation tools (GQS, mDISCERN, and MQ-VET) and their relationships with video feature metrics (Figure 4). The results indicated a highly significant positive correlation between the scores of GQS, mDISCERN, and MQ-VET, with p-values all less than 0.05. Regarding video feature associations, each evaluation tool exhibited distinct patterns: GQS showed a weak positive correlation with video length (r = 0.40, p < 0.001), number of likes (r = 0.31, p < 0.001), number of comments (r = 0.29, p < 0.001), as well as background environment (r = 0.26, p = 0.002), visual communication design (r = 0.23, p = 0.001), and Index Integrity (r = 0.20, p < 0.001). The mDISCERN score demonstrated a low positive correlation with video length (r = 0.31, p < 0.001), visual communication design (r = 0.24, p = 0.01), and Index Integrity (r = 0.17, p = 0.007). The MQ-VET score exhibited a moderate positive correlation with video length (r = 0.57, p < 0.001) and low positive correlations with the number of likes (r = 0.18, p = 0.005), number of saves (r = 0.13, p = 0.041), number of comments (r = 0.14, p = 0.027), background environment (r = 0.18, p = 0.005), visual communication design (r = 0.36, p < 0.001), and Index Integrity (r = 0.19, p = 0.004).

Figure 4.

The Spearman correlation coefficients between video general information and quality and reliability of video content. ×p > 0.05.

Keywords analysis

In this study, a keyword cloud analysis method was employed to visualize the video text data. The font size and centrality of the terms in the word cloud are positively correlated with their frequency of occurrence, meaning that high-frequency terms are displayed in a larger font size in the core area of the graph. Data analysis revealed that in the dimension of detection indicators, “WBC” (n = 199), “HGB” (n = 191), and “PLT” (n = 174) ranked as the top three in term frequency (Figure 5(a)). In the disease classification dimension, “anemia” (n = 176), “infectious diseases” (n = 175), and “coagulation-related diseases” (n = 151) were the three most frequently occurring categories of diseases (Figure 5(b)).

Figure 5.

Word cloud map. (A) Dimension of detection indicators; (B) Disease classification dimension.

Discussion

This study systematically evaluated metadata, engagement metrics, content quality, and reliability of 242 Chinese online videos related to CBC, uncovering potential issues in the dissemination of medical testing knowledge via new media. Although CBC is a foundational diagnostic tool widely utilized in clinical practice, the overall quality of related educational videos varies considerably. Notable disparities exist across platforms and creator identities, with some content posing a risk of misinformation that could adversely affect public health awareness.

Overall video evaluation

Our findings indicate that the quality of Chinese CBC-related online videos is moderate but marked by significant heterogeneity and critical shortcomings. The median GQS was 3.0, with only 49.5% of videos achieving a high-quality threshold (GQS≥4). The mDISCERN score, a measure of content reliability, also had a median of 3.0, suggesting average reliability, with just 28.9% of videos meeting the reliability standard (mDISCERN ≥ 4). The MQ-VET, a more comprehensive metric, yielded a median score of 58.0, indicating acceptable quality. However, specific subitems such as “recording date” and “resources and references” scored notably low, highlighting a need for improved reliability. Furthermore, video quality was significantly associated with the poster's identity, with CBC videos posted by medical practitioners outperforming those by nonmedical practitioners across all three quality metrics (p < 0.05).

Correlation analyses revealed a modest positive relationship between video length and quality, as longer videos can convey more health information and elucidate topics more thoroughly. ³⁰ However, no clear correlation emerged between video quality and engagement metrics such as “likes,” “comments,” and “shares,” aligning with prior studies on nonalcoholic fatty liver disease ²⁶ and liver cancer. ²² This suggests that the general public has limited capacity to evaluate the quality of health education videos. Intriguingly, a negative trend was observed between video length and engagement metrics, possibly due to viewers losing interest in longer videos. ³¹

The delivery design, visual background, and subtitle text constitute a three-dimensional audiovisual framework critical to the accurate comprehension and retention of medical testing knowledge. In our study, 74.38% of videos employed animation as a visual delivery method, and 76.86% included open captions, facilitating understanding of complex CBC content. Animation leverages rich graphics, vivid effects, and humor to captivate viewers, serving as an effective communication tool. ³² Subtitles enhance comprehension, attention, and memory retention, increasing appeal to learners. ³³ Additionally, 56.61% of videos used virtual backgrounds, 4.13% featured laboratory settings, and 26.03% utilized other hospital environments. Videos filmed in hospitals typically featured doctors as protagonists, with visible faces and narration in their own voices, fostering a stronger connection between learners, content, and presenters. ³⁴ In the remote context of online platforms, this “human touch” enhances the perceived presence of doctors, ³⁵ while hospital settings bolster viewer trust, improving receptivity to medical advice.

Regarding specific content, completeness was lacking: only 46.7% of videos referenced all CBC parameters, and 8.68% omitted specific indicators entirely. Word cloud analysis identified “WBC,” “HGB,” and “PLT” as the top three frequently mentioned indicators, reflecting their familiarity to the public. The most cited diseases—“Anemia,” “Infectious Diseases,” and “Coagulation Disorders”—showed strong associations with these indicators, consistent with clinical practice and public interest in prevalent conditions.

Concurrently, our study identified the proliferation of absolute assertions as the most prominent source of misinformation in CBC-related science communication videos on Chinese social media platforms. Exemplary claims include “CBC interpretation requires only three parameters” and “Elevated WBC count indicates bacterial infection.” These videos, predominantly produced by licensed physicians, frequently employ categorical narration and exaggerated sound effects, achieving broad dissemination through algorithmic amplification. Such diagnostic oversimplification yields multifaceted harms: it not only risks misleading patients into self-diagnosis that may delay clinical investigation of complex etiologies, but also readily triggers unnecessary health anxiety by disregarding interpretive complexities, including individual variations and benign causes of mild abnormalities. When discrepancies emerge between clinical diagnoses and viral online catchphrases, public skepticism toward healthcare professionals escalates, fostering uncritical adherence to fragmented information that disrupts standard clinical workflows. This phenomenon fundamentally reflects short-video platform dynamics, where creators prioritize engagement-driven simplification (e.g., “Elevated WBC = Bacterial Infection”) to accommodate information-compressed formats, perpetuating a cyclical pattern where virality reinforces oversimplification. Consequently, scientific accuracy becomes subordinated to dissemination efficiency. Furthermore, our research documented frequent instances of misleading anecdotal content regarding CBC interpretation. Nevertheless, such material does not constitute genuinely harmful misinformation or disinformation, consistent with JAMA's distinction: “It's not based on science; it's based on experience.” ³⁶

Platform differences and characteristics

Our study focused on three major platforms, Douyin, Rednote, and BiliBili whose characteristics aligned with our expectations. Below is a detailed analysis:

As a short video platform, Douyin relies on algorithms, ³⁷ offering abundant music and effects to aid content creation. Its videos are characterized by brevity, frequency, and speed, ³⁸ with evident content homogenization. The median time since publication on Douyin (63 days) was significantly shorter than on BiliBili (440 days) and Rednote (165 days), likely reflecting its algorithm's emphasis on recency. ³⁹ CBC video duration on Douyin (57 s) was markedly shorter than on BiliBili (704 s). As noted, content richness depends partly on length and information volume, with longer videos typically conveying more detail. Yet, even for medical videos, Douyin's traffic hinges on viral elements, ⁴⁰ necessitating a balance between quality and entertainment rather than scientific depth alone. Despite shallower content and brevity, Douyin leverages viral features (e.g., trending tags, suspenseful music) to capture attention, balancing warning effects with traffic generation, though often leaning toward superficiality and gimmickry.

Notably, Douyin enforces stringent identity verification, with CBC videos typically labeled “for reference only” or “seek offline care if unwell.” This may explain its higher quality scores compared to other platforms, particularly in MQ-VET's “identity of the presenter” item and mDISCERN's “additional sources listed” item. To post medical content, creators must submit credentials (e.g., ID, medical license, department proof) for committee review, earning a “Douyin Doctor Yellow V Certification” badge. Displayed publicly, this enhances trust. Certified doctors are barred from e-commerce, reducing conflicts of interest—a model other platforms could adopt. Nonetheless, even certified doctors incorporate viral elements to broaden their reach. ⁴¹

Rednote's certification mirrors Douyin's, requiring posters to be full-time doctors from top-tier hospitals or attending physicians from other public hospitals, with proof. This high entry barrier curbs profit-driven misinformation. Its strong social focus, driven by user-generated content (UGC), encourages sharing authentic experiences, including medical insights. Among 89 retrieved CBC videos, seven were patient perspectives, often of lower quality, potentially lowering Rednote's overall scores. This aligns with its social orientation.

BiliBili, popular among youth and students, ⁴² features medium-to-long videos, and slower commercialization. Its medical science videos are notably more professional, likely due to its audience of medical professionals and students. We categorized posters as licensed physicians or nonphysicians, given their differing CBC expertise. University professors and training institution instructors often provided in-depth analyses but were classified as science communicators due to nonpractitioner status. Uniquely, BiliBili acts as a “content porter,” ⁴³ with posters reposting CBC learning videos from public sources such as conference lectures. These have clear origins, deep content, and target professionals, explaining why medical practitioners’ median MQ-VET score (60) was lower than nonmedical practitioners’ (61). This diversity promotes knowledge sharing.

Selection and analysis of video evaluation tools

We used GQS, mDISCERN, and MQ-VET to assess CBC video quality and reliability. Although the JAMA score ²³ was considered, its overlap with MQ-VET led to its exclusion. GQS and mDISCERN, validated in fields such as nonalcoholic fatty liver disease, ²⁶ inflammatory bowel disease, ⁴⁴ and migraines, ²⁵ falter with low-information-density videos, prone to subjective bias. The MQ-VET scale, developed by Guler's team, ²⁵ offers a standardized framework (e.g., completeness, accuracy, utility), improving objectivity and reproducibility, as validated elsewhere. ²⁶

Global contextual considerations

Although our study focuses on Chinese social media platforms, it is undeniable that digital video as a novel modality for health communication is exerting global impact. YouTube, as the world's most widely used long-form video platform, has amassed substantial active users through multilingual support. While some studies indicate that medical content on YouTube remains relatively unvetted for factual accuracy with often unverifiable information, ⁴⁵ most research demonstrates its superior quality in health-related videos compared to China's domestic long-form platform Bilibili.^21,46 This advantage may correlate with YouTube's higher proportion of institutional accounts, which facilitates higher-quality content production, more frequent updates, and larger subscriber base development. ⁴⁷ However, YouTube remains inaccessible in China. ²¹ TikTok, as the international counterpart of Douyin under ByteDance ownership, similarly faces access restrictions in China. ⁴⁸ Multiple comparative analyses indicate that health-related content quality on Douyin surpasses TikTok,^49,50 likely attributable to Douyin's medical credential verification mechanisms. In contrast, TikTok's health videos predominantly originate from private users sharing personal symptom experiences and medical histories, resulting in lower efficacy and quality. To some extent, TikTok shares similarities with Rednote in platform positioning, where subjective UGC creates ambiguous zones. Regarding other platforms, Instagram and Twitter are generally considered inappropriate venues for sharing medical video content, ⁵¹ while Facebook serves as a viable but suboptimal health communication resource compared to YouTube. ⁵²

Furthermore, national-level regulatory frameworks play a pivotal guiding role in health communication. As early as 2023, China's Cyberspace Administration of China mandated platforms to verify medical content creators’ qualifications, requiring public display of credentials (service qualifications, professional certifications, expertise backgrounds) and affixing domain-specific labels on account profiles. ⁵³ The U.S. regulatory environment focuses predominantly on combating false advertising, fraudulent medical claims, and undisclosed conflicts of interest. The Food and Drug Administration monitors social media content promoting pharmaceuticals or medical devices for advertising regulation compliance, ⁵⁴ while the Federal Trade Commission regulates deceptive commercial practices—including influencer promotions of health products without disclosure. ⁵⁵ The European Union's recently implemented Digital Services Act represents an alternative regulatory approach, imposing heightened systemic obligations on Very Large Online Platforms. These platforms must conduct risk assessments (including public health risks) and implement corresponding mitigation measures, ⁵⁶ potentially triggering stricter content moderation or labeling mechanisms for sensitive domains like health. Ultimately, enhanced regulatory oversight of health information dissemination on social media proves crucial for facilitating stricter platform policies, curbing misinformation proliferation, and improving public discernment of online information.

Limitations

This study has limitations. Data were confined to Chinese platforms, excluding other languages and limiting cross-linguistic comparisons. Current tools, while robust (e.g., MQ-VET), struggle to capture professional depth, with some BiliBili videos exceeding 5-point scales. Dimensions such as “cutting-edge” or “interdisciplinary integration” are absent, hindering quantification of academic value.

Practical significance and future recommendations

Accurate CBC interpretation is vital for public health. To enhance dissemination quality, platforms should verify professional creators, display credentials, and prioritize evidence-based content via algorithms. For complex terms, pop-up definitions or animations could reduce cognitive load. Additionally, we propose translating select international health videos where copyright permits. Finally, we urge national regulators to implement proactive policies imposing reputational sanctions on accounts/platforms promoting pseudoscience. These recommendations target medical-health video governance globally, extending beyond China. Future research could explore multilingual platforms and cultural impacts, developing tiered tools to assess basic and advanced content, shifting platforms from “traffic-first” to “quality-first” for a trusted CBC knowledge ecosystem, ultimately providing sustainable pathways for advancing population-wide health literacy.

Conclusion

This pioneering study elucidates the dissemination and quality of CBC-related videos on BiliBili, Douyin, and Rednote, revealing deficiencies (GQS: 3.0, mDISCERN: 3.0, MQ-VET: 58.0). Platform traits shape quality: Douyin excels but favors shallow warnings; BiliBili's professional focus yields polarized quality; Rednote's patient videos risk bias, underscoring public interpretive limits.